Long-running Transactions Research Articles

Revisiting software transactional memory in Haskell

Software Transactional Memory (STM) has become very popular in Haskell. Currently, there are nearly 500 packages on Haskell’s package archive that directly use STM. Despite the widespread use in real world applications, Haskell’s STM implementation has seen very few updates since its introduction in 2005. In this work, we describe our efforts to redesign the STM implementation in the Glasgow Haskell Compiler (GHC), based on a TL2-like implementation that is able to support both <code>orElse</code> and <code>retry</code> without the use of traditional nested transactions. We argue that our implementation is simpler than the current GHC implementation while supporting opacity. We also demonstrate that our implementation performs better than the current GHC implementation on a number of benchmarks by multiple orders of magnitude for long-running transactions. In an effort to measure the performance of <code>orElse</code> and <code>retry</code>, we present an STM-based work stealing scheduler. With <code>orElse</code> and <code>retry</code>, we are able to elegantly implement the scheduler in just a few lines of code. We have modified the Par Monad, a real-world Haskell package that provides deterministic parallelism, to use our STM-based work stealing scheduler and show that it is not only simpler but is able to perform as well as the current scheduler.

Modeling of Compensation in Long-Running Transactions

nowadays, the most controversial issue is transaction in database systems or web services. Specifically, in the area of service-oriented computing, where business transactions always need long periods of time to finish. In the case of a failure rollback, which is the traditional method, it will not be enough and not suitable for handling errors during long running transactions. As a substitute, the most appropriate approach is compensation which is used as an error recovery mechanism. Therefore, transactions that need a long time to complete are programmed as a composition of a set of compensable transactions. This study attempts to design several compensation policies in the long running web transaction especially when the transaction has parallel threads. Meabwhile, one thread in sequence steps of the transaction may fail. This paper also describes and models many different ways to compensate to the thread. Moreover, this study proposes a system to implement creating long running transactions as well as simulating failures by using compensation policies.

True Concurrency in Long-running Transactions for Digital Ecosystems

The concept of a digital ecosystem (DE) has been used to explore scenarios in which multiple online services and resources can be accessed by users without there being a single point of control, which can be used to effectively serialise their interactions. We argue in this paper that this weak coupling between services places additional demands on the modelling of compensation and recovery management in long-running transactions over traditional SOC related formalisms. We describe an adaptation of Shields’ vector languages, in that the synchronisation constraint is removed (no shared actions), as a formal semantics for a transaction in terms of the common ordering constraints on the underlying interactions between its participants. The notation afforded by the so-called transaction languages captures the invocations on each participant service (online resource), and at each point during execution, across the whole transaction. Concurrency is modelled explicitly through a notion of independence, which is lifted onto tuples of sequences (one for each participant of the transaction) rather than individual sequences, as in Mazurkiewicz trace languages or events, as in the event structures model. Participating subcomponents execute concurrently and failure of one or more causes the recovery of the whole transaction. Compensations are triggered immediately upon failure and concurrent forward actions are compensated concurrently. We highlight the benefits of our true-concurrent approach in the context of DEs and outline connections of transaction languages to other partial order models. Further, we discuss how our approach supports forward recovery in that recovering the whole transaction is avoided wherever possible.

Failure Recovery Dependency Analysis for Web Composition Transactions Based on Extended Petri-Nets

For transactional Composition Web Services (CWS), traditional transaction theory in the database is introduced into the Long Running Transactions (LRTs) domain. In practical application, the typical LRTs employ forward or backward rollback. In order to improve the expansibility of LRTs, it is important to accurately specify the dependency between tasks in composition transaction; contemporary technologies usually statically specify dependency point and avoid implicit interaction in LRTs. In this paper, we propose a task dependency modeling method for LRTs, firstly we present formal description of atomic Web Services (WS) and discuss atomic WS such as pivot, compensable, retriable and vital, then execution logs of LRTs data dependency and action dependency rules between tasks in LRTs are analyzed, specify, finally application of TRP shows that it is feasible to ensure consistent execution of reliable LRTs based dependencies.

Adaptive transaction scheduling for mixed transactional workloads

Optimistic scheduling policy based on the concept of Very Important Transaction.Progressively pessimistic scheduling policy with low runtime overheads.Adaptive scheduler that combines two scheduling policies in a synergistic way.Thorough evaluation of 4 schedulers with the STAMP and the STMBench7 benchmark. Transaction schedulers reduce the number of transaction reexecutions in applications using Software Transactional Memory (STM) by preventing conflicting transactions to run in parallel. Unfortunately, current scheduling solutions are too conservative, rely on coarse measures to serialize transactions, and are specially inadequate for workloads with long transactions.In this paper we introduce an optimistic and adaptive transaction scheduler that takes advantage of the information already collected by the STM runtime to increase parallelism between transactions and, thus, improve transactions' throughput. Our new ProVIT scheduler implements a low-overhead scheduling policy for short transactions that reduces concurrency as contention increases and a fine-grained scheduling policy for long transactions based on the novel concept of Very Important Transaction.Experimental results conducted with the STMBench7 benchmark and the STAMP benchmark suite showed that the ProVIT scheduler has comparable performance to other current scheduling solutions for short transactions, but is up to 65% faster for long-running transactions.

CJoin: Join with communicating transactions

This paper proposes a formal approach to the design and programming of long running transactions (LRTs). We exploit techniques from process calculi to define cJoin, which is an extension of the Join calculus with few well-disciplined primitives for LRT. Transactions in cJoin are intended to describe the transactional interaction of several partners, under the assumption that any partner executing a transaction may communicate only with other transactional partners. In such case, the transactions run by any party are bound to achieve the same outcome (i.e., all succeed or all fail). Hence, a distinguishing feature of cJoin, called dynamic joinability, is that ongoing transactions can be merged to complete their tasks and when this happens either all succeed or all abort. Additionally, cJoin is based on compensations i.e., partial executions of transactions are recovered by executing user-defined programs instead of providing automatic rollback. The expressiveness and generality of cJoin is demonstrated by many examples addressing common programming patterns. The mathematical foundation is accompanied by a prototype language implementation, which is an extension of the JoCaml compiler.

On the behaviour of general purpose applications on cloud storages

Managing data over cloud infrastructures raises novel challenges with respect to existing and well-studied approaches such as ACID and long-running transactions. One of the main requirements is to provide availability and partition tolerance in a scenario with replicas and distributed control. This comes at the price of a weaker consistency, usually called eventual consistency. These weak memory models have proved to be suitable in a number of scenarios, such as the analysis of large data with map reduce. However, due to the widespread availability of cloud infrastructures, weak storages are used not only by specialised applications but also by general purpose applications. We provide a formal approach, based on process calculi, to reason about the behaviour of programs that rely on cloud stores. For instance, it allows to check that the composition of a process with a cloud store ensures ‘strong’ properties through a wise usage of asynchronous message-passing; in this case, we say that the process supports the consistency level provided by the cloud store. The proposed approach is compositional: the support of a consistency level is preserved by parallel composition when the preorder used to compare process-store ensembles is the weak simulation.

Recovery within long-running transactions

As computer systems continue to grow in complexity, the possibility of failure increases. At the same time, the increase in computer system pervasiveness in day-to-day activities bring along increased expectations on their reliability. This has led to the need for effective and automatic error-recovery techniques to resolve failures. Transactions enable the handling of failure propagation over concurrent systems due to dependencies, restoring the system to the point before the failure occurred. However, in various settings, especially when interacting with the real world, reversal is not possible. The notion of compensations has been long advocated as a way of addressing this issue, through the specification of activities which can be executed to undo partial transactions. Still, there is no accepted standard theory; the literature offers a plethora of distinct formalisms and approaches. In this survey, we review the compensations from a theoretical point of view by (i) giving a historic account of the evolution of compensating transactions; (ii) delineating and describing a number of design options involved; (iii) presenting a number of formalisms found in the literature, exposing similarities and differences; (iv) comparing formal notions of compensation correctness; (v) giving insights regarding the application of compensations in practice; and (vi) discussing current and future research trends in the area.

Compensation by design

Abstract The current dominance of the service-based paradigm reflects the success of specific design and architectural principles embodied in terms like SOA and REST. This paper suggests further principles for the design of services exhibiting long-running transactions (that is, transactions whose characteristic feature is that in the case of failure not all system states can be automatically restored: system compensation is required). The principles are expressed at the level of scope-based compensation and fault handling, and ensure the consistency of data critical to the business logic. They do so by demanding (a) either the commitment of all of the transaction or none of it, and (b) that compensation is assured in case of failure in ‘parent’ transactions. The notion of scope is captured algebraically (rather than semantically) in order to express design guidelines which ensure that a given transaction satisfies those principles. Transactional processes are constructed by parallel composition of services, and transactions with scopes in a single service are dealt with as a special case. The system semantics is formalised as a transition system (in Z) and the principles are expressed as formulae in linear temporal logic over runs of the transition system. That facilitates the model checking (using SAL) of their bounded versions. Two simple examples are used throughout to illustrate definitions and finally to demonstrate the approach.

Failure-divergence semantics and refinement of long running transactions

Compensating CSP (cCSP) models long-running transactions. It can be used to specify service orchestrations written in programming languages like WS-BPEL. However, the original cCSP does not allow to model internal (non-deterministic) choice, synchronized parallel composition, hiding or recursion. In this paper, we introduce these operators and define for the extended language a failure-divergence (FD) semantics to allow reasoning about non-determinism, deadlock and livelock. Furthermore, we develop a refinement calculus that allows us to compare the level of non-determinism between long running transactions, and transform specifications for design and analysis.

A Comprehensive Failure Recovery Algorithm for LRTs Based on Paired Net

Failure recovery optimization is an important way for enhancing efficiency of Long Running Transaction (LRT) processing. In this paper, to solve the efficiency problem of LRT failure recovery, a Comprehensive Recovery Model for LRTs (LCRM) is constructed, which divides LRTs into a series of sub-transactions in different levels, and supports versatile transaction properties of LRT. Based on LCRM, a Comprehensive Failure Recovery algorithm for LRTs (LCFR) is proposed. This algorithm uses methods of forward recovery, backward recovery and alternative recovery. It supports auto-recovery of failures during the execution of LRTs. LCFR guarantees LRT’s semantic atomicity property and durability property. By restricting the recovery scope in lower level of complex LRTs, LCFR limits the quantity of sub-transactions to be recovered. Thus, it reduces unnecessary loss of time and enhances the efficiency of failure recovery. Experiment results show that LCFR can reduce the time required for failure recovery and decrease the failure rate of LRT processing.

High-performance concurrency control mechanisms for main-memory databases

A database system optimized for in-memory storage can support much higher transaction rates than current systems. However, standard concurrency control methods used today do not scale to the high transaction rates achievable by such systems. In this paper we introduce two efficient concurrency control methods specifically designed for main-memory databases. Both use multiversioning to isolate read-only transactions from updates but differ in how atomicity is ensured: one is optimistic and one is pessimistic. To avoid expensive context switching, transactions never block during normal processing but they may have to wait before commit to ensure correct serialization ordering. We also implemented a main-memory optimized version of single-version locking. Experimental results show that while single-version locking works well when transactions are short and contention is low performance degrades under more demanding conditions. The multiversion schemes have higher overhead but are much less sensitive to hotspots and the presence of long-running transactions.

Formal Design and Verification of Long-Running Transactions with Extensible Coordination Tools

Ensuring transactional behavior of business processes and web service compositions is an essential issue in the area of service-oriented computing. Transactions in this context may require long periods of time to complete and must be managed using nonblocking techniques. Data integrity in long-running transactions (LRTs) is preserved using compensations, that is, activities explicitly programmed to eliminate the effects of a process terminated by a user or that failed to complete due to another reason. In this paper, we present a framework for behavioral modeling of business processes, focusing on their transactional properties. Our solution is based on the channel-based coordination language Reo, which is an expressive, compositional, and semantically precise design language admitting formal reasoning. The operational semantics of Reo is given by constraint automata (CA). We illustrate how Reo can be used for modeling termination and compensation handling in a number of commonly used workflow patterns, including sequential and parallel compositions, nested transactions, discriminator choice and concurrent flows with link dependences. Furthermore, we show how essential properties of LRTs can be expressed in LTL and CTL-like logics and verified using model checking technology. Our framework is supported by a number of Eclipse plug-ins that provides facilities for modeling, animation, and verification of LRTs to generate executable code for them.

Static vs Dynamic SAGAs

SAGAs calculi (or simply SAGAs) have been proposed by Bruni et al. as a model for long-running transactions. The approach therein can be considered static, while a dynamic approach has been proposed by Lanese and Zavattaro. In this paper we first extend both static SAGAs (in the centralized interruption policy) and dynamic SAGAs to deal with nesting, then we compare the two approaches.

Transaction Scripts: Making Implicit Scenarios Explicit

We describe a true-concurrent approach for managing dependencies between distributed and concurrent coordinator components of a long-running transaction. In previous work we have described how interactions specified in a scenario can be translated into a tuples-based behavioural description, namely vector languages. In this paper we show how reasoning against order-theoretic properties of such languages can reveal missing behaviours which are not explicitly described in the scenario but are still possible. Our approach supports the gradual refinement of scenarios of interaction into a complete set of behaviours that includes all desirable orderings of execution and prohibits emergent behaviour of the transaction.

On the expressive power of process interruption and compensation

The investigation into the foundational aspects of linguistic mechanisms for programming long-running transactions (such as thescopeoperator of WS-BPEL) has recently renewed the interest in process algebraic operators that, due to the occurrence of a failure,interruptthe execution of one process, replacing it with another one called thefailure handler. We investigate the decidability of termination problems for two simple fragments of CCS (one with recursion and one with replication) extended with one of two such operators, theinterruptoperator of CSP and thetry-catchoperator for exception handling. More precisely, we consider the existential termination problem (existence of one terminated computation) and the universal termination problem (all computations terminate). We prove that, as far as the decidability of the considered problems is concerned, under replication there is no difference between interrupt and try-catch (universal termination is decidable while existential termination is not), while under recursion this is not the case (existential termination is undecidable while universal termination is decidable only for interrupt). As a consequence of our undecidability results, we show the existence of an expressiveness gap between a fragment of CCS and its extension with either the interrupt or the try-catch operator.

Security in a Model for Long-running Transactions

Communicating Hierarchical Transaction-based Timed Automata have been introduced to model systems performing long-running transactions. Here, for these automata a security concept is introduced, which is based on a notion of opacity and on the assumption that an attacker can not only observe public system activities, but also cause abortion of some of them. Different intruder capabilities as well as different kinds of opacity are defined and the resulting security properties are investigated. Security of long-running transactions is defined by the mentioned notion of opacity and conditions for compositionality are established.

Design and verification of long-running transactions in a timed framework

Long-running transactions consist of tasks which may be executed sequentially and in parallel, may contain subtasks, and may require to be completed before a deadline. These transactions are not atomic and, in case of executions which cannot be completed, a compensation mechanism must be provided. In this paper we develop a model of Communicating Hierarchical Timed Automata suitable to describe the mentioned aspects in a framework where also time is taken into account. We develop the patterns for composing long-running transactions sequentially, in parallel or by nesting. The correct compensation of a composed long-running transaction is preserved by these composition patterns. The automaton-theoretic approach allows the verification of properties by model checking. As a case study, we model and analyse an example of e-commerce application described in terms of long-running transactions.

The Pervasive Workflow: A Decentralized Workflow System Supporting Long-Running Transactions

<para xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> Workflow technologies are becoming pervasive in that they enable the execution of business processes in distributed and ubiquitous computing environments. As long-running transactions, the execution of workflows in environments without dedicated infrastructures raises transactional requirements due to the dynamicity of resources available to run a workflow instance and the integration of relaxed atomicity constraints at both design and instantiation time. In this paper, we propose an adaptive transactional protocol for the pervasive workflow model developed in a previous work to support the execution of business processes in the pervasive setting. The execution of this protocol takes place in two phases. First, candidate business partners are assigned to tasks using an algorithm wherein the selection process is based on both functional and transactional requirements. The workflow execution further proceeds through a hierarchical coordination protocol managed by the workflow initiator and controlled based on a decision table computed as an outcome of the business partner assignment procedure. The resulting workflow execution is compliant with the defined consistency requirements, and the coordination decisions depend on the transactional characteristics offered by the partners assigned to each task. An implementation of our theoretical results relying on ontology Web Language for Series and Business Process Execution Language technologies is further detailed as a proof of concept. </para>

A Formal Framework for Web Services Coordination

Recently the term Web Services Choreography has been introduced to address some issues related to Web Services Composition and Coordination. Several proposals for describing Choreography for Business Processes have been presented in the last years and many of these languages (e.g. BPEL4WS) make use of concepts as long-running transactions and compensations for coping with error handling. However, the complexity of BPEL4WS makes it difficult to formally define this framework, thus limiting the formal reasoning about the designed applications. In this paper, we formally address Web Services Coordination with particular attention to Web transactions. We enhance our past work - the Event Calculus - introducing two main novelties: i) a multicast event notification mechanism, and ii) event scope names binding. The former enables an easier specification of complex coordination scenarios — such as E-commerce applications require — while the latter allows many new interesting behaviors which can be very useful in business scenarios: the introduction of private event scope names — used to deal with security and privacy — and a dynamic event scopes definition that can be used to manage multiple instances of the same application.